Effects of Adversarial Patches Against Object Detection on Combating Cheats in Online First Person Shooter Games

In FPS (First Person Shooter) games, there exist problematic cheats called “aimbots” that automate a player's (i.e., cheater's) aiming actions at enemies. Such cheats have become a social problem because they cause game companies to lose credibility and users to leave the games. With advances in CV (Computer Vision) technology, CV-based cheat tools for FPS games are becoming pervasive. CV-based cheats are more difficult to be detected than cheats that read in-game memory because they use CV to get enemy's positions. This paper proposes a method to avoid CV-based cheats by applying an adversarial patch to each enemy character, and validates the effects of the proposed method on rate of avoiding YOLOv5-based object detection with respect to loR (Intersection over Recognized object's bounding box) instead of IoU (Intersection over Union) for thresholds of mAP (mean-Average Precision).